AppFlowy MCP
An MCP server for AppFlowy Cloud that enables workspace, space, page, trash, favorite, and page-content management.
README
AppFlowy MCP
An MCP server for AppFlowy. This fork adds workspace folder, space, page, trash, favorite, and basic page-content tools on top of the original workspace/database/row tools.
Requirements
- Python 3.14
- uv
- AppFlowy account credentials
Configure In Codex
codex mcp add appflowy -- uvx appflowy-mcp --env APPFLOWY_EMAIL="your-email" --env APPFLOWY_PASSWORD="your-password"
Option 1: PyPI / uvx
Use this after the package is published to PyPI. This is the simplest setup for daily use.
Use uvx as the command:
uvx
Use these arguments:
appflowy-mcp
Option 2: Local Source / uv
Use this when developing the MCP locally or running directly from a cloned repository.
Use uv as the command:
uv
Use these arguments:
run
--project
D:\codes\MCP\Appflowy-MCP
D:\codes\MCP\Appflowy-MCP\main.py
Configure In Claude-code
claude mcp add appflowy -- uvx appflowy -e APPFLOWY_EMAIL=your-email -e APPFLOWY_PASSWORD=your-password
Environment Variables
Set these environment variables:
APPFLOWY_EMAIL=your-email@example.com
APPFLOWY_PASSWORD=your-password
With these variables set, tools automatically log in on first use. appflowy_login is still available when you want to provide credentials explicitly.
For self-hosted AppFlowy, also set the base URL:
APPFLOWY_BASE_URL=http://localhost:8000
If omitted, APPFLOWY_BASE_URL defaults to https://beta.appflowy.cloud.
AppFlowy Structure
Most write operations need both a workspace and a parent view:
workspace -> space -> page/database
To create a page inside a space, pass the workspace ID and use the space view_id as parent_view_id.
Tools
Authentication
appflowy_loginappflowy_refresh_token
Workspaces And Spaces
appflowy_list_workspacesappflowy_get_workspace_folderappflowy_list_spacesappflowy_create_spaceappflowy_update_space
Pages
appflowy_create_pageappflowy_get_pageappflowy_update_pageappflowy_move_page_to_trashappflowy_restore_page_from_trashappflowy_delete_page_from_trashappflowy_favorite_pageappflowy_list_trashappflowy_list_favorites
Page Content
appflowy_save_pageappflowy_append_page_contentappflowy_append_text_to_pageappflowy_append_blocks_to_pageappflowy_create_markdown_pageappflowy_append_markdown_to_pageappflowy_import_markdown_fileappflowy_import_markdown_directory
Page-content support covers appending new document blocks, creating pages from Markdown, and appending Markdown to existing pages. AppFlowy exposes a high-level append-block endpoint, but not a matching high-level REST endpoint for deleting or editing arbitrary existing blocks. Page-level deletion through trash is supported.
When saving AI answers, notes, summaries, or generated content into AppFlowy, use appflowy_save_page by default. Its content_format field defaults to markdown, so agents do not need to ask users to say "save as Markdown" every time. Set content_format to plain_text only when the user explicitly wants the content preserved as literal plain text.
Use appflowy_append_page_content by default when adding AI-generated content to an existing page. It also treats content as Markdown unless content_format is set to plain_text.
Example paragraph block:
{
"type": "paragraph",
"data": {
"delta": [
{
"insert": "Hello from MCP"
}
]
}
}
Example Markdown page:
{
"parent_view_id": "space-or-page-view-id",
"title": "Meeting Notes",
"content": "# Meeting Notes\n\n- [ ] Follow up\n- **Important** decision\n\n```python\nprint(\"hello\")\n```",
"content_format": "markdown"
}
Markdown conversion supports headings, paragraphs, dividers, bullet lists, numbered lists, todo lists, quotes, code blocks, image links, and inline bold, italic, strikethrough, code, and links.
Markdown Import
Use appflowy_import_markdown_file to import one local Markdown file as an AppFlowy page. Use appflowy_import_markdown_directory to recursively import a local folder tree.
Directory import maps local structure directly to AppFlowy:
local folder -> AppFlowy page
local subfolder -> AppFlowy subpage
local .md/.markdown file -> AppFlowy page
README.md or index.md becomes the content of its folder page instead of a separate child page. Other Markdown files in the same folder become child pages. Hidden folders plus .git, .hg, .svn, .idea, .vscode, node_modules, and __pycache__ are skipped.
Local image references are resolved relative to the Markdown file, uploaded to AppFlowy file storage, and replaced with AppFlowy file URLs:


Remote image URLs such as https://... are kept as-is. Missing local images are reported in the import result warnings, while other files continue importing.
Example directory import:
{
"parent_view_id": "space-or-page-view-id",
"path": "D:\\notes",
"upload_assets": true
}
Example single-file import:
{
"parent_view_id": "space-or-page-view-id",
"path": "D:\\notes\\MCP\\config.md",
"title": "MCP Config",
"upload_assets": true
}
Databases And Rows
appflowy_list_databasesappflowy_get_database_fieldsappflowy_list_rowsappflowy_get_row_detailsappflowy_create_rowappflowy_upsert_rowappflowy_get_updated_rows
Local Run
uv run appflowy-mcp
Publish To PyPI
Build and check the package:
uv build
uv publish --dry-run --trusted-publishing never
Publish with a PyPI API token:
$env:UV_PUBLISH_TOKEN="pypi-your-token"
uv publish --trusted-publishing never
Do not commit PyPI tokens or write them into project files.
Notes
- Tokens are stored in memory by the MCP server process.
APPFLOWY_EMAIL,APPFLOWY_PASSWORD, andAPPFLOWY_BASE_URLcan also be provided through a local.envfile.- Some page and space endpoints are implemented from AppFlowy source routes that are not present in the public OpenAPI document.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。